I have a close friend working in compliance for a major bank and they do use multiple software to detect risky transactions and clients.

Flagging is nothing, it would flag based on name similarity and all kinds of stuff but a human would have to review it. They use scoring, review clients and transactions while prioritizing on that score. Half of their time goes on optimization of the scoring and detection as the human review process is resource intensive.

If they had a system rising millions of flags, the first thing they would do is to reduce these flags to a manageable level. There’s no way that so many transactions are high risk, they would get rid of clients rising that many red flags if those are legitimately risky transactions.

If they don’t do good job in the compliance department, they receive multiple billion dollar fines. Simply, no client is valuable enough to risk these fines. Rise some flags and you’re out and reported to a regulator, they’re not going to let you flood the system with questionable transactions.